Feature Report on the clearing trial of Pteronia incana (bluebush) and Elytroppappus rhinocerotis (renosterbos) in Mgwalana catchment, Ngqushwa District, Eastern Cape Province: rehabilitating degraded rangeland

AR PALMER Agricultural Research Council, Animal Production Institute E mail: [email protected]

Introduction control measure. In order to evaluate the effective- nvasion of blue bush ( Pteronia incana ) and renos- ness and cost of clearing these species, a project Iterbos ( Elytropappus rhinocerotis ) continues in was initiated to experimentally clear a number of trial the rangeland of the Mgwalana catchment of the areas as part of the National Department of Agri- Ngqushwa district (formerly Peddie) of the Eastern culture’s Landcare Programme. Five sites of known Cape (Kakembo et al. 2006, Kakembo et al. 2007). bluebush and renosterbos infestation were identified These species, both indigenous to the Nama karoo from high resolution near infra-red (NIR) photography region of South Africa, successfully invade natural in the Mgwalana catchment. In four sites, intensive grassland and thicket. Both species have been re- manual clearing was carried out to assess the cost corded in un-disturbed rangelands across the Great and efficacy of three clearing techniques. At the fifth Fish river and Keiskamma river valleys, but are known site, a landscape-scale clearing action occurred to to increase and become dominant in the absence of measure the impact of clearing on the MODIS leaf fire or when excessive soil disturbance has occurred. area index (LAI). The sites varied in the density of Disturbance of the soil surface by excessive tram- infestation. pling, cultivation (Kakembo 2001) or subterranean mammal activity (e.g. termites or mole rats) creates Aim gaps which can be colonized by seedlings. The aim of this trial was to evaluate the cost and effectiveness of manual clearing of blue bush (laven- Control of both species on free-hold rangeland der bush) ( Pteronia incana ) and renosterbos ( Elytro- with adequate infra-structure can be affected using pappus rhinocerotis ) on the communal rangelands of fire. However, communal rangelands do not lend the Ngqushwa district. themselves to the use of fire, as there are no options for the control of grazing after burning, and little op- Methods portunity to build up fuel loads which would enable Three physical clearing treatments were applied at a burn to be applied. This is principally due to the each of the intensive clearing site in March 2003. absence of livestock control systems (adequately Treatments were applied to three randomly selected fenced and gated camps, water provision) in areas sub-plots of the five quadrangular sub-plots of ap- under communal tenure. The project aims at evaluat- proximately 3ha. (100mx300m). One sub-plot was ing the cost effectiveness of a range of manual clear- used as a control, and the fifth was used as the ing techniques which may be used as an alternative area to be packed with dead material and burned

22 GRASSROOTS | The GRASSLAND SOCIETY of SOUTHERN AFRICA | www.grassland.org.za August 2010 | Vol 10 No.3 Feature

at a later stage. A burn was applied to these sites on 17 September 2003. The three physical clearing treatments were applied during a single operation in March 2003. The first used a specially designed instrument or “dicker” (manufactured by Fabcomp, King Williams Town), which pulls the rooted plant out of the ground. The second used commercial lopping shears to cut the plant off at the base, and the third employed a “slasher” to chop the plant off.

At all sites, the density of large plants (>0.5m in height), was determined using the point-centred quarter (PCQ) technique. The boundaries of each treatment were determined from the aerial photo- graphs, and marked in the field using a GPS. Each team was allocated an equal amount of cash to com- plete the three treatments and pack the burn plot. The limiting factor to completing the 3 ha was the fi- nances allocated to each team. Teams were encour- aged to complete the full 3ha, but at the highest den- sities of infestation this was not possible within the financial allocation. At the completion of the 22 day working period, the area completed by each team was recorded using a GPS and the area calculated using a GIS package.

MODIS LAI product (Collection 4) We acquired 8 years of the MODIS 8-day 1 km Col- lection 4 (MOD15A2) composite LAI/FPAR product (Knyazikhin et al . 1998) for the period March 2000 to Photo: Tony Palmer. The dicker instrument being operated by December 2007 from the NASA Distributed Active one of the ladies from the village of Nyaniso, Archive Centre. The data were extracted from the archive using the MODIS re-projection tool (HegTool) and imported into IDRISI (Version I32.11, Clark Labs) image processing package to create an 8-year data stack. The geo-codes for the centre of site 5 and the two control sites (village, un-cleared and grassland) were entered into the GIS and 8-day LAI values for each site were extracted.

August 2010 | Vol 10 No.3 www.grassland.org.za | The GRASSLAND SOCIETY of SOUTHERN AFRICA | GRASSROOTS 23 Feature Rehabilitating Degraded Rangeland

Results The results of the costs associated with the various treatments are presented (Table 1).

Total person Treatment Site 1 Density Team Celetyuma days days Team Area (ha) worked Area/day size Cost per ha. Dicker 1.99441 High 38 0.052484 3 114 R 2,958.86 Lopper 2.359212 Low 7 0.33703 3 21 R 460.77 Slasher 2.971566 Low 6 0.495261 3 18 R 313.56 Pack 1.944959 17 0.114409 1 17 R 452.45 Control 3 Total 170 R 51.76

Site 2 Team Mgwalana days Team Area (ha) worked Area/day size Cost per ha. Dicker 0.985735 High 18 0.054763 3 54 R 2,398.39 Lopper 0.716058 High 21 0.034098 3 63 R 3,851.93 Slasher 0.78571 High 21 0.037415 3 63 R 3,510.47 Pack 0.135282 21 0.006442 1 21 R 6,796.21 Control 0.764488 Total 201 R 43.78 Team Site 3 Mgwalana:Youth days Team Area (ha) worked Area/day size Cost per ha. Dicker 0.322315 High 22 0.014651 3 66 R 8,190.74 Lopper 0.547779 High 22 0.024899 3 66 R 4,819.46 Slasher 0.649184 High 22 0.029508 3 66 R 4,066.65 Pack 0.62375 22 0.028352 1 22 R 1,410.82 Control 0.759029 Total 220 R 40.00

Site 4 Team:Amahlubi days Team Area (ha) worked Area/day size Cost per ha. Dicker 2.429 Moderate 11 0.220818 3 33 R 934.03 Lopper 2.409477 Moderate 14 0.172105 3 42 R 1,198.39 Slasher 2.870326 Moderate 13 0.220794 3 39 R 934.13 Pack 0.52434 14 0.037453 1 14 R 1,835.64 Control 2.575065 Total 128 R 68.75

24 GRASSROOTS | The GRASSLAND SOCIETY of SOUTHERN AFRICA | www.grassland.org.za August 2010 | Vol 10 No.3 Feature

Near infra-red imagery was recorded in October 2004 over 2 sites using the Kodak DCS420 infra- red digital camera. Eighteen months after clearing (October 2004) there were some visible differences in the near infra red reflectance and NDVI’s (Figure 1) between treatments, but these were not all signifi- cant and their interpretation must be dealt with cau- tiously. The dicker treatment on Site 3 had a higher infra-red signal and NDVI than the other treatments (Figure 1).

Mean Plot (Spreadsheet2 10v*44c) 0.20

0.18

0.16

0.14

0.12

0.10

0.08

NDVI_2004 0.06

0.04

0.02

0.00

-0.02

-0.04 Mean 1 2 3 4 5 6 7 Mean±0.95*SE Treatment Figure 1. NDVI values for the 4 treatments, control, adjacent woody thicket and good condition grassland in October 2004. NDVI was calculated from the NIR and red bands of the Kodak DCS420 digital camera. 1=Dicker, 2=Control, 3=Lopper, 4=Slasher, 5=Burned (Falkia repens absent)/ bare soil, 6=Riparian thicket (woody species), 7=Good grassland, un-invaded by Pteronia incana.

This pattern was not repeated at site 2. This may be due to the stimulation of the soil microbial activ- ity after complete extraction of plants by the dicker, resulting in the release of nutrients and greater active green growth. During October 2004, these burn scars were covered by a single herbaceous species, Falkia repens. This was not a desirable species, and as there was only one herbaceous species present, this was likely to die off during the dry season, leaving no ground cover. The NDVI of woody shrubs and healthy grassland were significantly higher than Pteronia incana stands and clearing treatments (Figure 2).

August 2010 | Vol 10 No.3 www.grassland.org.za | The GRASSLAND SOCIETY of SOUTHERN AFRICA | GRASSROOTS 25 Feature Rehabilitating Degraded Rangeland

Mean Plot NDVI_2001 0.12

0.10

0.08

0.06

0.04

0.02

0.00

NDVI -0.02

-0.04

-0.06

-0.08

-0.10

-0.12

-0.14 Mean 1 2 3 4 5 Mean±0.95*SE Treatment

Figure 2. The mean and standard errors of the DCS 420 NDVI’s for the five treatments. The riparian zone (Treatment 4), with its woody component and high perennial grass cover has the highest NDVI, followed by the control grassland in good condition (Treatment 2). 1= sparse grassland with Pteronia incana, 2=grassland with Elytropappas rhinocoerotis (control), 3=Pteronia incana shrubland, 4=riparian thicket, 5=sparse shrubland on shallow rocky soil.

During a site visit in June 2007, it was clear that there was an increase cover of bluebush from those seedlings that had not been extracted during the clearing operation. The clearing techniques only removed large plants and left many seedlings and seeds. Opening up of the gap created space for the seedlings to grow and they have been growing vigor- ously. I expect a higher NDVI for these sites during 2008, as the emerging bluebush seedlings are in an active growth phase. Further assessment was planned during 2008, with the possible purchase of more ASTER imagery. Using a supervised classifica- tion procedure on the ASTER data, we defined the area invaded by high densities of Pteronia incana to be approximately 27000ha. The image did not cover the whole of Ngqushwa, and therefore represents an under-estimate of the extent of the problem. A time series plot of the 8-day MODIS leaf area index (LAI) of the cleared area and the three control sites was provided for the period April 2000 to December 2007.

Prior to the clearing event in November 2003, the Pteronia infested site (Site 5) had a lower LAI than

26 GRASSROOTS | The GRASSLAND SOCIETY of SOUTHERN AFRICA | www.grassland.org.za August 2010 | Vol 10 No.3 Feature the grassland site and a higher NDVI than the village. approach, an area of 30ha was cleared during No- This matched well with the expectation for the site, vember 2003. This area (Figure 4), comprising a large as the 1km MODIS pixel included some good con- proportion of high density infestation, was cleared at dition grassland. After clearing in November 2003, a cost of R1000 per ha. This represents an example there was a decline in the LAI for that site for the of how the results of the clearing trials can be applied whole of 2004 compared with the control-village and to large scale clearing and rehabilitation of range- control grassland where, no treatment took place. land. Results on the kill rate of the three clearing This agrees with the reduction in leaf area that would techniques showed that loppers and slashers did not have accompanied the clearing of 30ha of P. incana achieve a significant kill rate (<5%). Dickers achieve at site 5. The decline in MODIS LAI was significant at a high kill rate (100%) of adult plants but there was a the end of 2004, but in the growing season of 2005 substantial seedling re-growth. Follow up is required. the MODIS LAI had increased to a value midway The burns were too hot, caused soil scorching and between the control-village and control grassland. were colonized by a single species. This was not a This was an indication of the extent of the re-growth desirable state and hot burning (scorching) should of P. incana after the treatment. The very low LAI be avoided at all costs. However a more appropri- values of the treated-site (Site 5), control-village and ate burning programme could be considered if a rest control- P. incana relative to the control-grassland can be guaranteed after the burn. This would have to are indicative of extremely low production poten- include the cost of fencing to exclude all herbivores tial of the invaded grasslands. In order to quantify for a lengthy period (6-12 months) before and after the possible impact of this difference on grassland the burn. production, I integrated the area under the two time series plots for the cleared and grassland sites. References Kakembo V, 2001. Trends in Vegetation Degradation in The leaf area index of the undisturbed grassland Relation to Land Tenure, Rainfall, and Population in the growing season was significantly higher than Changes in Peddie District, Eastern Cape, South that of the degraded rangeland at the other three Africa Environmental Management Vol. 28, No. sites (village, cleared and control- P. incana ). This dif- 1, pp. 39–46. ference can be attributed to the lower water use ef- ficiency of the senescent stands of P. incana and the Kakembo V, Palmer AR, Rowntree K. 2006. The use effectiveness of grassland in using rainfall to capture of high resolution digital imagery to characterize carbon. A simple integration of the area under the the distribution of Pteronia incana invaser curve in a normal year (e.g. 2004) reveals that the species in Ngqushwa (formerly Peddie) District, annual LAI of the grassland exceeds that of the de- Eastern Cape, South Africa. International Journal graded P. incana by 213%. In addition, P. incana was of Remote Sensing 27: 2735-2752. totally un-available to livestock, so this figure could be inflated by perhaps as much as another 200%, Kakembo V, Rowntree, K and Palmer AR 2007. making the production of the grassland nearly 5 Topographic controls on the invasion of Pteronia times greater than that of the degraded rangeland. incana (Blue bush) onto hillslopes in Ngqushwa (formerly Peddie) district, Eastern Cape, South Conclusion Africa. Catena 70 (2007) 185–199. The project provided an indication of the approximate labour cost of clearing varying densities of bluebush Knyazikhin Y, J. V. Martonchik, R. B. Myneni, D. J. Diner, infestation. Under very dense conditions, labour and S. W. Running. 1998. Synergistic algorithm costs per hectare cleared were extremely high (Table for estimating vegetation canopy leaf area index 1). These densities did not prevail at consistent levels and fraction of absorbed photosynthetically throughout the landscape, and an approach was active radiation from MODIS and MISR data. recommended where groups are invited to quote on Journal of Geophysical Research-Atmospheres clearing large, previously measured areas. Using this 103:32257-32275.

August 2010 | Vol 10 No.3 www.grassland.org.za | The GRASSLAND SOCIETY of SOUTHERN AFRICA | GRASSROOTS 27